Weekly Articles
Clear guidance on choosing sustainable technology, building the right product, and connecting business requirements to maintainable engineering systems.
More Writing
Alongside this site, I publish practical articles on AI workflows, cloud architecture, automation, QA, product engineering, and technology leadership.
Local LLM
What I learned running coding, reasoning, and agentic AI models on an RTX 3060 12GB, and why model orchestration became the harder problem.
AI Agents
Poor AI agent performance is rarely solved by a bigger model. Real enterprise results come from context infrastructure that captures, organizes, and delivers the right project, module, and code context to any agent.
Coding Agents
A practical architecture for giving coding agents the right code, project knowledge, and memory without filling every prompt with unnecessary context.
RAG
How retrieval-augmented generation works best when chunking, hybrid retrieval, and reranking are tuned to the document type: transcripts, contracts, specs, status reports, and user stories—plus how ContextHarbor bakes those profiles in.
AI Home Lab
Why I built a hybrid AI home lab where local and cloud models work together as one engineering platform.
AI Home Lab
How GGUF Switchboard brings workload scheduling, memory-pressure eviction, and automatic model lifecycle management to local llama.cpp inference.
AI Home Lab
How Portainer, OmniRoute, Beszel, and Uptime Kuma turn a collection of local AI services into a platform I can operate reliably.
AI Home Lab
How a private Local Project RAG platform gives AI assistants persistent, searchable context across engineering projects without relying on cloud services.
AI Product Development
A practical workflow for designing and building digitization products with process workshops, mind maps, user stories, architecture, AI coding agents, MCP tools, and build discipline.
Forward Deployed Engineer
How embedding a single engineer end-to-end eliminates translation loss and ships better enterprise AI applications — from requirements through design thinking, architecture, and production.
AI Strategy
Every operational business wants AI. Most are not ready for it. The problem is almost never the technology — it is what comes before it.
Technology Strategy
Tech stacks age. If yours is slowing your team, killing your hiring, or making business decisions for you, it is time to take a hard look.
Fractional CTO
Both write code. Only one is thinking about your business. Here is what separates a developer from a CTO — and why it matters before you hire.
Fractional CTO
Why business leaders need a bridge from real requirements to proven, maintainable technology choices.
AI Strategy
A pragmatic way for operational businesses to adopt AI without creating brittle systems or expensive experiments.
LangChain
Should you build your LangChain service in Python or integrate it into a NestJS ecosystem? A comparison across performance, maintainability, hiring, and speed.
OpenAI
A scalable, cloud-native resume skill-matching bot built with Google Apps Script, Google Sheets, and OpenAI, turning Sheets into a lightweight ATS.
LangChain
How I used LangChain and Pinecone to turn a reactive chat assistant into a remembering, context-aware product, with lessons from production.
Kong
Deploy Kong as an API gateway in front of microservices using Docker Compose and Deck declarative config, with a bonus GitHub Actions workflow for Cloud Run.
Tailwind CSS
After 48 hours of debugging broken layouts and purged classes in a Next.js app, here's what went wrong with Tailwind v4 and how we rolled back to v3.3.5.
Firebase
Step-by-step guide to integrating Microsoft login with Firebase Auth in a Next.js + TypeScript app, including Azure/Entra setup, code, and troubleshooting.
QA Automation
Why moving QA automation from Java to Python delivered a major boost in development speed, plus the Python testing ecosystem worth knowing.
Next.js
After three days of debugging file uploads in a Next.js API route, the fix turned out to be five lines that disable the default bodyParser.
OSM
A guide to parsing and analysing OSM PBF data using Python and Osmium, extracting node/way/relation counts, way distances, and intersection stats.
Fine Tuning
Why general-purpose LLMs fall short on domain-specific tasks, how fine-tuning fixes it, and the open-source LLMTrainingTools repo for building Q&A datasets.